265

Pilot-scale struvite recovery from anaerobicdigester supernatant at an enhanced biologicalphosphorus removal wastewater treatment plant

A. Britton, F.A. Koch, D.S. Mavinic, A. Adnan, W.K. Oldham, and B. Udala

Abstract: Pilot testing of a fluidized bed reactor used to recover phosphate, in the form of struvite, from anaerobicdigester supernatant was conducted at the Advanced Wastewater Treatment Plant, City of Penticton, British Columbia. Themain objective of this study was to demonstrate the ability of the reactor to remove at least 70% of the phosphate in thesupernatant from a digester fed with a combination of primary and secondary sludge. It was found that the operation ofthe reactor could be controlled to achieve any desired level of phosphorus removal up to 90%. Analysis of the recoveredstruvite crystals showed essentially pure struvite (>99% by weight) with small amounts of calcium (<0.5% by weight)and traces of potassium and iron. The recovered crystals had mean diameters increasing from 0.5 to 1.8 mm over thecourse of the study. The increasing diameters are believed to be the result of changes in the crystal structures that causedthem to become stronger over the course of the study.

Key words: crystallization, nutrient removal, phosphorus recovery, struvite, sludge treatment, wastewater.

Résumé : Des essais pilotes d’un réacteur à lit fluidisé utilisé pour la récupération du phosphate, sous forme de struvite, àpartir d’un surnageant d’un digesteur anaérobie, ont été effectués sur place, dans l’usine « Advanced Wastewater TreatmentPlant », située dans la ville de Penticton, en Colombie-Britannique. L’objectif principal de l’étude était de démontrer lacapacité du réacteur à éliminer au moins 70 % du phosphate contenu dans le surnageant d’un digesteur alimenté d’unecombinaison de boues primaires et secondaires. Nous avons trouvé que l’opération du réacteur pourrait être contrôlée afind’atteindre un niveau de récupération du phosphate pouvant atteindre 90 %. L’analyse des cristaux de struvite récupérésa montré que la struvite était essentiellement pure (>99 % poids) avec de petites quantités de calcium (<0,5 % poids)et des traces de potassium et de fer. Les cristaux récupérés avaient des diamètres moyens augmentant de 0,5 à 1,8 mmdurant l’étude. Cette augmentation de diamètre est probablement causée par des changements dans la structure cristalline,solidifiant les cristaux au fur et à mesure de la progression de l’étude.

Mots clés : cristallisation, retrait des nutriments, récupération du phosphore, struvite, traitement des boues, eaux usées.

[Traduit par la Rédaction]

Introduction

Phosphorus recovery is an important area of research in theenvironmental engineering field. There are a variety of reasonsfor this, including the gradual depletion of global reserves ofhigh-quality mined phosphate deposits and operationalproblems encountered in wastewater treatment plants (Nied-

Received 8 March 2004. Revision accepted 1 September 2004.Published on the NRC Research Press Web site at http://jees.nrc.ca/on 16 July 2005.

A. Britton,1 F.A. Koch,2 D.S. Mavinic, and A. Adnan. Depart-ment of Civil Engineering, University of British Columbia, Van-couver, BC V6T 1Z4, Canada.W.K. Oldham. Stantec Consulting Ltd., #1007-7445 132nd Str.,Surrey, BC V3W 1J8, Canada.B. Udala. City of PentictonAdvanced Wastewater Treatment Plant,Penticton, BC, Canada.

Written discussion of this article is welcomed and will be receivedby the Editor until 30 November 2005.

1 Present Address: Ostara Nutrient Recovery Technologies, Inc.,#690-1199 West Pender Str., Vancouver, BC V6E 2R1, Canada.2 Corresponding author (e-mail: koch@civil.ubc.ca).

bala 1995; Jardin and Popel 1996; Driver et al. 1999). The pip-ing and equipment in the sludge treatment processes of theseplants are increasingly prone to fouling and encrustation withstruvite, thereby increasing pumping and maintenance costs.Another important factor driving this research is the increasedopportunity for phosphate recovery derived from the increaseduse of enhanced biological phosphorus removal (EBPR). Theuse of EBPR in wastewater treatment leads to the creation ofan enriched phosphate stream in the sludge-handling liquors(Woods et al. 1999). This enriched stream has been the focusof most investigations into the recovery of phosphorus frommunicipal wastewaters.

In this study, a pilot-scale phosphorus recovery reactor, de-veloped at The University of British Columbia (UBC), wastested at the Advanced Wastewater Treatment Plant (AWWTP)in Penticton, British Columbia. This reactor was used to recoverphosphate in the form of struvite (MgNH4PO4·6H20) from ananaerobic digester supernatant stream, through the addition ofmagnesium chloride and pH adjustment in a fluidized bed. Thestudy was carried out over a 4-month period, during which timethe anaerobic digester was fed with a blend of primary and sec-ondary sludge.

J. Environ. Eng. Sci. 4: 265–277 (2005) doi: 10.1139/S04-059 © 2005 NRC Canada

266 J. Environ. Eng. Sci. Vol. 4, 2005

Site selectionThe City of Penticton AWWTP is one of the best-performing

cold-climate EBPR plants in North America, and the staff hasbeen extremely helpful in previous collaborative research con-ducted on site. The City of Penticton has also expressed aninterest in being one of the pioneering municipalities in Can-ada with regards to implementing innovative and sustainablewaste management strategies.

The AWWTP consists of preliminary treatment (includingscreening and degritting), primary clarifiers, two parallel mod-ified University of Cape Town design EBPR secondary treat-ment trains, sand filters for secondary effluent polishing, andchlorine disinfection by chlorination and dechlorination. TheCity of Penticton is located in the Okanagan Valley in south-central British Columbia, between Lake Okanagan and LakeSkaha. This is a region of intense horticulture and crop irriga-tion and an environment sensitive to excessive nutrient input.For this reason, the wastewater treatment plant is subject tovery low effluent phosphorus standards (0.25 mg/L PO4-P).The sludge treatment train for this plant consists of primarysludge fermenters that provide the required volatile fatty acidsfor the EBPR system and a two-stage anaerobic digester forthe fermented primary sludge. The digested sludge is then de-watered on a belt press after being combined with thickenedwaste-activated sludge (WAS) from the EBPR trains. The WASis not digested on site because this practice leads to the releaseof the excess phosphorus in the sludge from the EBPR system(Niedbala 1995; Mavinic et al. 1998). Instead, the combineddewatered sludge is composted off site in windrows at the mu-nicipal landfill, and the composted sludge is sold as a soil condi-tioner to local landscaping and agricultural operations. Duringthe course of this study, the operation of the sludge treatmentsystem was modified to transfer a portion of the thickened WASto the digester to obtain a digester supernatant stream rich inphosphate and ammonia.

Research objectivesThe main purpose of this study was to use real supernatant

to test the reactor at a full-scale EBPR wastewater treatmentplant. Previous research at UBC had successfully demonstrateda reactor design used to recover struvite from synthetic super-natants with Mg, NH4-N, and PO4-P concentrations similar tothose expected in the digester supernatant at the City of Pen-ticton AWWTP; however, some concerns remained as to theeffects of dissolved constituents and suspended solids in thereal supernatant on the control and operation of the reactor.Both bench- and pilot-scale experiments had been conductedat UBC prior to this on-site study, and the expected operatingparameters were determined prior to setting up the pilot plantat Penticton (Dastur 2001; Adnan et al. 2003a).

The prime objective of this study was to determine the oper-ational parameters that would allow successful operation of thepilot-scale reactors treating real anaerobic digester supernatantfrom a full-scale EBPR wastewater treatment plant. Success-ful operation was defined as the controlled removal of at least

Fig. 1. Pilot-scale struvite crystallizer reactor process design.

Supernatant

storage

2 X 16 m3

MgCl

1 m3

NaOH

1 m3

202 mm ID

77 mm ID

52 mm ID

MAP crystallizer

Effluent

External

clarifier40 mm ID

harvest

section

Injection

port

Recycle

Sludge

pH

control

probe

70% of the orthophosphate from the digester supernatant andthe recovery of this phosphorus in the form of large (>1 mm)and easily separable struvite crystals.

Materials and methods

Reactor design, process description, and operationBased on previous experiments at the bench scale, a pilot-

scale reactor was designed and tested at the UBC Environmen-tal Engineering Pilot Plant using a synthetic feed (Adnan etal. 2003a). Two identical reactors based on this design wereinstalled and operated in parallel over a 4-month period, fromSeptember to December 2001, at the City of PentictonAWWTP.They were housed in an on-site, heated chemical storage build-ing.

Figure 1 shows the basic design of the reactor and associ-ated equipment. The reactor itself was a fluidized bed reactorwith sections of increasing diameter and a settling zone at thetop. The diameter changes caused turbulent eddies above eachtransition, ensuring that sufficient mixing existed in the reactor,and also helped to classify the fluidized particles by size. Onlythe largest crystals in the reactor were harvested.

The crystallizer was constructed of clear PVC piping con-nected with standard fittings. The inside diameters were 40, 52,and 77 mm, respectively, for the bottom, middle, and top sec-tions of the fluidized zone. The clarifier section at the top ofthe crystallizer was built out of 202-mm-diameter clear acrylicpipe. The total liquid volume of each reactor was approximately19 L, 9 L of which were in the three fluidized zones. The totalheight of the reactors was approximately 4900 mm. The totalliquid flow rate through each reactor for the duration of thestudy was 3.6 L/min. Each reactor was equipped with two pHprobes, one in the top of the harvest zone and another in the ex-ternal clarifier. The main function of the external clarifier wasto recycle the effluent back into the reactor. It was also used to

© 2005 NRC Canada

Britton et al. 267

trap the washed-out fine crystals from the reactor. A stainlesssteel injection port was provided at the base of the reactor toblend the influent, recycle, and chemical-dosing streams.

The pH probe in the harvest zone, with a proportional flowpH controller, was used for feedback control. Magnesium wasdosed to the reactor in the form of magnesium chloride solutionto supply the desired magnesium to phosphorus molar ratio inthe reactor.

Chemicals, storage tanks, and pumps

The digester at the City of Penticton AWWTP is a two-stageanaerobic digester, operated in the mesophilic (36−40 ◦C) tem-perature range. The first stage of the digester is gas mixed; thesecond stage is unmixed. Supernatant from the second stage ofthe digester was pumped from an overflow splitter box into twostorage tanks for use in the pilot struvite crystallizer. Each tankhad a capacity of 16 000 L and was equipped with overflowsand drain valves. The supernatant was pumped out of the tanksfrom a fitting located approximately 500 mm above the tankbottom; this allowed any suspended solids in the supernatantto settle to the bottom of the tank and prevented excess solidsfrom entering the reactor. One full tank would typically lastapproximately 7 d when feeding both reactors.

The large volume of the storage tanks also allowed a constantfeed strength to be maintained for several days at a time. Thiswas crucial during the initial months of the study, when the su-pernatant characteristics were changing significantly on a dailybasis. Feed supernatant was pumped from the storage tanks us-ing a Moyno™ model 500 331 progressive cavity pump, witha 1/2 HP AC motor and variable frequency drive, to allow pre-cise flow control (capacity 0.3–7.6 L/min). Each reactor hadits own independent pumping system to ensure accurate flowcontrol and to allow different types of operation to be evaluatedin parallel.

Because the solubility of struvite is highly pH dependent(Abe 1995; Doyle et al. 2000), a pH control system is critical tomaintain the desired supersaturation conditions in the crystalliz-ers. For this study, the pH was adjusted using sodium hydroxide(caustic soda). The pH in the reactor was monitored at the topof the harvest zone (as shown in Fig. 1) using a Cole-Parmerdouble-junction, in-line pH probe. The pH was controlled basedon this signal using a Cole-Parmer model EW-56025-40 pHcontrol system with proportional output.

Magnesium was found to be the limiting reagent in the forma-tion of struvite from the Penticton digester supernatant; thus,it was necessary to supplement the reactor with magnesium.Throughout this study, the objective was to keep a molar ratio ofMg:PO4-P equal to 1.3:1 within the reactor. This higher magne-sium concentration causes the limiting reagent to be phosphateand thus allows for lower effluent phosphate concentrationsthan in magnesium-limited systems. A Mg:PO4-P ratio of 1.3:1was previously found to reliably provide low-effluent phospho-rus concentrations while allowing large crystals to be grownin previous bench-and pilot-scale studies (Dastur 2001; Adnan

2002). In this study, the magnesium requirements were met us-ing a solution of magnesium chloride hexahydrate.

Sampling and analysesSamples of the digester supernatant and the effluent from

each reactor were collected daily and subsequently filtered andanalyzed for total magnesium or [Mg+2]total, total ammonia–nitrogen or [NH4-N]total, and total orthophosphate or [PO4-P]total. All samples were filtered using Fisherbrand G6 filterpapers with a nominal pore size of 1.5µm to remove suspendedsolids from the samples. [Mg+2]total was analyzed using flameatomic absorption spectrophotometry (model Varian Inc. Spec-trAA220™ fast sequential atomic absorption spectrophotome-ter). Calcium, aluminum, iron, and potassium analyses wereperformed on the samples from the crystal product analysis.These analyses were also performed by atomic absorption spec-trophotometry. On-site ammonia tests were performed with aHach DR2000 spectrophotometer, using the salicylate method(Hach, Nitrogen,Ammonia, High RangeTest’NTube™ method10031). Orthophosphate was measured on site using the stan-nous chloride method, as described in method number 4500-P D in Standard Methods for the Examination of Water andWastewater (APHA/AWWA/WEF 1995), with the exceptionthat sample sizes were 50 mL instead of 100 mL. Sample ab-sorbances were measured using a Milton Roy Spectronic 401spectrophotometer.

Crystal harvest procedureCrystals were harvested from the reactor after the feed, re-

cycle, and chemical feed flows were stopped, and the crystalswere allowed to settle. The harvest zone (see Fig. 1) was iso-lated using ball valves after the settling was complete, and thesettled bed volume of struvite crystals was measured. To re-move the crystals for harvesting, the injection-port section ofthe reactor was removed and the crystals were allowed to fallinto a bucket. The harvest zone was then rinsed with reactoreffluent to ensure that all crystals were removed. The harvestedcrystals were then air dried before analysis.

Crystal product analysesSamples of several of the harvested crystals were dissolved

in a 0.5% nitric acid solution to determine the composition andpurity of the crystals grown in the reactor. These solutions weresubsequently analyzed for the components of struvite, as wellas calcium, aluminum, iron, and potassium. For each sampleanalyzed, a known weight of struvite crystals (approximately30 mg) was dissolved in 50 mL of 0.5% nitric acid solution. Toaccelerate the crystal dissolution, the samples were submergedin an ultrasonic bath until no visible solids remained. Sampleswere then inverted several times and vortex mixed before anal-ysis.

Struvite solubility determinationThe solubility product or Ksp, as defined in this study, is the

product of the ionic activities of the precise ionic forms involved

© 2005 NRC Canada

268 J. Environ. Eng. Sci. Vol. 4, 2005

in the formation of a precipitate. For struvite, this relation isdefined by eq. [1], where the { } braces indicate ion activity inmoles per litre.

[1] Ksp = {Mg2+}{

NH4+}{

PO43−}

The problems associated with using Ksp for reactor con-trol has been discussed elsewhere (Britton 2002; Adnan et al.2003b). A more practical way to operate the reactor is to use theconcept of conditional solubility product (Snoeyink and Jenk-ins 1980; Ohlinger 1999; Britton 2002; Adnan et al. 2003b).The struvite conditional solubility product (Ps), as employedin this study, is the direct product of the analytical results forsoluble magnesium, ammonia nitrogen, and orthophosphate, asdefined by eq. [2], where the [ ] brackets indicate concentrationin moles per litre.

[2] Ps = [Mg2+]

total

[NH4 − N

]total

[PO4 − P

]total

An equilibrium conditional solubility curve was developedusing the digester supernatant from the City of PentictonAWWTP, over a range of pH values. This curve was devel-oped for use as a control parameter for struvite crystallizationat the pilot scale (Britton 2002).

ApparatusThe apparatus used for determining the solubility of struvite

was a six-station paddle stirrer (Phipps and Bird). Square jarscontaining 1.5 L of the solution being tested were immersed ina constant temperature bath at 20 ± 0.1 ◦C. The paddle stirrerswere set to operate at 70 ± 2 rpm. A sufficient mass of struvitecrystals, harvested from the reactors, was placed in each jar toensure that some solid-phase struvite remained at equilibrium.Equilibrium was assumed to be reached 24 h after conditionswere changed in each jar, as shown by previous research atUBC (Adnan et al. 2003a). The pH in each jar was adjustedusing dilute hydrochloric acid and sodium hydroxide solutionsto determine the solubility of struvite over the expected oper-ating range of struvite crystallization equipment (i.e., betweenpH 7 and 9). After 24 h for a given set of conditions, the pHand conductivity in each jar were measured. Conductivity wasmeasured using a Hanna Instruments HI9033 multi-range con-ductivity meter. Samples were filtered using 0.45-µm What-man filter paper, preserved, and later analyzed for magnesium,ammonia-nitrogen, and orthophosphate, according to the ana-lytical methods described earlier. It should be noted that for the24-h equilibrium period, it was assumed that the struvite waspure and crystal size did not affect struvite solubility to anyextent.

TerminologyFeed supersaturation ratio

The feed or influent supersaturation ratio (SSR) describesthe hypothetical SSR that would exist in an instantly mixedsolution consisting of digester supernatant, magnesium chlo-ride, and sodium hydroxide, in proportions equal to those fed

to the reactor. The SSR was calculated based on eq. [3], wherePS sample represents the calculated conditional solubility prod-uct of the sample being analysed and PS equilibrium represents theequilibrium conditional solubility product for that same sample,based on an equilibrium conditional solubility curve developedas described above.

[3] SSR = PS sample/PS equilibrium

In-reactor supersaturation ratioThe SSR in the reactor was calculated using the same method-

ology as the feed SSR; however, in this case PS sample is calcu-lated by combining the concentrations of magnesium, ammo-nia, and orthophosphate in the reactor feed and recycle streams.Therefore, it is representative of the SSR in a completely mixedsample drawn from immediately above the injection ports ofthe reactors.

Recycle ratioThe recycle ratio represents the ratio of the flow from the

recycle pump to the combined flow from the supernatant feedpump and chemical-dosing pumps. This recycle ratio is usedto control the in-reactor SSR by diluting the feed with treatedeffluent. Details on this method of controlling reactor operationsare provided elsewhere (Britton 2002; Adnan et al. 2003b).

Crystal retention timeThe concept of crystal retention time (CRT) was developed

by the UBC project team to quantify the crystal age (i.e., theaverage time the crystals spend in the reactor). Analogous to theconcept of solids retention time in an activated sludge process,details on CRT calculation can be found elsewhere (Britton2002; Adnan et al. 2003a).

Mean crystal sizeThe mean crystal size of each harvest was calculated by siev-

ing and weighing each size fraction of the dried struvite crystalproduct. All the crystals in each size fraction were assumed tobe of a diameter in the middle of the size range. For example,the crystals that were <0.5 mm were assumed to be 0.25 mmin diameter, the 0.5–1-mm crystals were assumed to have adiameter of 0.75 mm, and so on.

Percent phosphate removalBecause the removal of phosphorus was a primary objective

of this research, the percentage of phosphate removed fromthe digester supernatant stream was monitored. This was calcu-lated based on a mass balance on phosphorus around the reactor(Britton 2002).

Results and discussion

The reactor design described in this paper was successful inrecovering phosphate in the form of struvite from a full-scaledigester supernatant at the City of Penticton AWWTP. Crys-talline product was recovered from the reactor as small pellets,with average diameters approaching 2 mm by the end of the

© 2005 NRC Canada

Britton et al. 269

Fig. 2. Struvite pPS in digester supernatant and distilled water as a function of pH.

study. These crystals were found to be nearly pure struvite andof a hardness adequate to allow easy separation and processingof the product.

The overall operation period of the reactors was from2 September to 13 December 2001. However, because of thelow phosphate (PO4-P) content (7.8–18.8 mg/L) in the di-gester supernatant initially, the operation of the reactors before12 October 2001 was essentially a commissioning phase andlittle phosphorus recovery was possible. During this time, highchemical dosages of both magnesium chloride and sodium hy-droxide were necessary to induce the crystallization of struvite.These early data are not presented here, other than to say thatphosphate removal to <5 mg/L by struvite crystallization ispossible but at a high operating cost.

Struvite solubility product determinationSeveral authors have attempted to determine a solubility prod-

uct for struvite, but there is a wide range of reported solubilityvalues (Dastur 2001). Therefore, it was considered importantfor this study to determine the equilibrium conditions that couldbe expected, when treating real digester supernatant from thePenticton AWWTP.

Thermodynamically, there should be a single value of Kspthat should apply to all solutions, as long as it is possible todetermine the activity of each chemical species accurately. Un-fortunately, this is difficult to do for individual compounds indigester supernatant because of the presence of a myriad ofknown and unknown compounds. The presence of these com-pounds also leads to a wide range of possible competing reac-tions, which could skew the solubility product determination.

Struvite conditional solubility productAs a simple means of determining the saturation state of the

supernatant being treated, a well-defined concept of PS wasused (Ohlinger 1999; Britton 2002; Adnan et al. 2003b). Fig-ure 2 shows the experimentally determined struvite PS curvesfor digester supernatant. A second order polynomial curve wasfitted to the data using Microsoft Excel software, and this curvewas used subsequently to represent equilibrium conditions inthe solutions. For the supernatant curve, eq. [4] describes this

polynomial curve, where pPS is the negative logarithm of thePS. This curve fits the data with a R2 value of 0.993, indicat-ing that this is an accurate representation of the equilibriumconditions in this particular supernatant.

[4] pPs = −0.203pH2 + 4.09pH − 11.76

Supernatant characteristics during the studyThe operation of the digester at the City of PentictonAWWTP

was modified during the course of this study. This resulted insignificant changes to the composition of the supernatant fromthe digester. Normal operation of the digester involved only thedigestion of primary sludge, resulting in PO4-P concentrationsof 5–15 mg/L in the digester supernatant.At the beginning of thestudy (early September), the digester was supplemented withthickened WAS from a thickener tank. This practice appearedto hydraulically overload the digester, resulting in suspendedsolids concentrations of up to 2000 mg/L in the supernatant,and it was therefore discontinued until a better solution couldbe found.

Following this period, a method of transferring WAS from thegravity belt thickener was devised, thereby allowing the transferof much thicker sludge (approximately 5% solids). This prac-tice allowed much more WAS to be transferred to the digesterwithout hydraulic overloading, and the phosphate concentra-tion could increase without high suspended solids in the super-natant. Once this practice was established in early October, itwas possible to maintain much higher phosphate concentrationsin the digester, as can be seen in Fig. 3. The concentrations ofammonia and phosphate increased steadily as the WAS contentin the digester was increased. However, the magnesium con-centration remained relatively constant and independent of theammonia and phosphate concentrations; this is contrary to thebelief that magnesium is released in conjunction with phos-phate when EBPR sludge is digested anaerobically (Doyle etal. 2000; Jardin and Popel 2001). The reasons for this differenttrend were not investigated as part of this study, but they areunder investigation in follow-up work.

From 12 October to 13 December, the concentration ofPO4-P ranged from 37 to 71 mg/L, whereas the NH4-N concen-

© 2005 NRC Canada

270 J. Environ. Eng. Sci. Vol. 4, 2005

Fig. 3. Digester supernatant composition.

0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0

80.0

Date

PO

-Pand

Mg

concentr

ations

(4

mg/L

)

0

50

100

150

200

250

300

350

400

450

500

NH

-Nconcentr

ation

(4

mg/L

)

PO -P4 Mg NH -N4

Fig. 4. Percentage phosphate removal for each reactor.

0.00

10.00

20.00

30.00

40.00

50.00

60.00

70.00

80.00

90.00

100.00

Date

%P

hosphate

rem

oval

Reactor A Reactor B

tration ranged from 197 to 436 mg/L and the Mg concentrationranged from 11 to 35 mg/L. Because of the wide variation in thecomposition of the supernatant being fed to the crystallizationreactors, the operation of both reactors was continually modi-fied as needed to maintain stable crystal growth conditions inthe reactor.

Reactor operationIn general, the reactors operated as expected and in a manner

similar to the operation observed during previous trials withsimilar, pilot-scale reactors using synthetic supernatant (Adnanet al. 2003a). The crystals harvested from the reactor were gen-erally of a darker colour and of a smaller diameter than thoseharvested from synthetic supernatant; however, removal andrecovery of these crystals were easy and the product-handlingmethods used in the previous trials worked well here.

Phosphate removal efficiencyOne of the main objectives of this research was to remove

phosphate from the digester supernatant stream in a full-scale

EBPR plant. Overall, it was possible to control the phospho-rus removal efficiency within the range of 30% to 90%. Thiscontrol was exerted either by setting the pH in the reactors orby setting the inlet SSR. The original objective of the studywas to demonstrate that it was possible to remove at least 70%of the orthophosphate from the digester supernatant. Figure 4presents the orthophosphate removal efficiencies achieved dur-ing the study, demonstrating adequate removal control.

Variations from the target value were investigated tocompare the economics of different removal efficiencies. Thetwo reactors were operated in a parallel until 7 November,when the operation of reactor B was modified to purposelyachieve lower orthophosphate removal. This was done by re-ducing pH, which lowered the inlet SSR, while maintaining thesame Mg:PO4-P molar ratio as reactor A. With the exception of1 d, when the magnesium feed to the reactor was accidentallyinterrupted (24 October), the removal of orthophosphate in reac-tor A was maintained above 70%. This shows that it is possibleto maintain the removal efficiency targeted in this study and thatit was possible to maintain a higher removal efficiency if desired

© 2005 NRC Canada

Britton et al. 271

Fig. 5. Phosphate removal vs. operating pH in the struvite crystallizing reactors.

0.00

10.00

20.00

30.00

40.00

50.00

60.00

70.00

80.00

90.00

100.00

pH

%P

hosphate

rem

oval

Reactor A Reactor B

Fig. 6. Phosphate removal vs. inlet supersaturation ratio.

0.00

10.00

20.00

30.00

40.00

50.00

60.00

70.00

80.00

90.00

100.00

Inlet SS Ratio

%P

hosphate

rem

oval

Reactor A Reactor B

(>80%) for a supernatant with an orthophosphate concentra-tion of at least 40 mg/L. These results are consistent with resultsobtained in several other studies where struvite was being re-covered from full-scale digester liquors (Abe 1995; Munch andBarr 2001; Ueno and Fujii 2001).

For orthophosphate removal, two possible factors were eval-uated to control the percentage removal: the operating pH ofthe reactor and the inlet SSR. The first assumes that the vari-ation in the Mg:NH4-N:PO4-P molar ratios in the supernatantis small and can therefore be ignored; also, the percentage ofphosphorus removed will vary with the operating pH, simplybecause the solubility of struvite, as defined by the equilibriumPS, varies with pH. This method of evaluation is by far the sim-plest and is useful for the day-to-day operation of the reactor;however, it would probably fail to predict accurately the per-formance of a reactor at a new site with different molar ratios.This relation is shown graphically in Fig. 5. The wide range of

removal efficiencies for a given pH (up to 31%) is due to thechange in inlet concentrations and changes in the Mg:NH4:PO4molar ratios during this study. In other words, this method ofprediction is useful, but simplistic, and would not be adequatefor a supernatant that is highly variable in composition if precisecontrol of the orthophosphate removal is required.

In the second method of predicting the removal of orthophos-phate, the inlet SSR is used. This method assumes that the efflu-ent is at equilibrium. Therefore, it is possible to predict the ef-fluent phosphate concentration by assuming equimolar removalof Mg, NH4-N, and PO4-P. Figure 6 shows the percentage ofphosphorus removal versus the inlet SSR for the operating pe-riod in both reactors. The important factors that contribute tothe scatter in Fig. 6 are the inaccuracies in the measurement ofthe Mg, NH4-N, and PO4-P concentrations, as well as in themeasurement of the pH. The fact that a SSR is a function of allfour of these measured values makes for a compounded error,

© 2005 NRC Canada

272 J. Environ. Eng. Sci. Vol. 4, 2005

Table 1. Comparison of theoretical struvite production basedon cumulative daily orthophosphate removal and actual struviterecovery based on cumulative struvite harvest weights during the3-month study period.

Reactor A Reactor B

Cumulative struvite harvested (kg) 7.82 6.50Struvite left in reactor at end of study (kg) 3.80 4.35Total struvite recovered (kg) 11.62 10.85Theoretical struvite produced (kg) 13.54 12.52Struvite recovered (%) 86 87

but the advantage is that this method allows the prediction of re-moval efficiency in a widely varying supernatant composition,with the same accuracy.

Overall, the best method to use to predict phosphate removalfrom an operational point of view will depend on the degree ofaccuracy wanted and the available data. A reactor can be con-trolled to remove the desired amount of phosphate by varyingthe operating pH or the inlet SSR. The difference between thetwo methods is that the first takes into account only one of thefactors involved and may be applicable only for a specific su-pernatant, while the latter requires a more complete analysis ofthe situation and is more generally applicable.

Struvite recoveryTo ensure that the phosphate being removed from the super-

natant was being recovered, the dry weight of each harvest ofstruvite was recorded, and the final dry weight of struvite ineach reactor was recorded at the end of the experiment. Thesemasses were compared with the theoretical mass of struvite thatshould have been formed, based on the phosphate removed fromthe digester supernatant (see Table 1). The struvite consideredto be recovered is the mass weighed after the drying and siev-ing of the harvested product. Some losses occurred during theprocess of harvesting, drying, transferring, sieving, and weigh-ing the product struvite crystals. Another source of error is thatthis analysis assumes that the reactors were operating 24 h aday, when in fact they were shut down daily to harvest crystals,monitor collapsed bed depth of crystals, clean injector ports,and calibrate pH probes. On average, these shutdown periodswere estimated to be 1 h per day or 4% of the time.

Table 1 shows that between 86% and 87% of the phosphateremoved was recovered. Correcting these values for the esti-mated shutdown time brings the theoretical recovery rates tobetween 90% and 91%. If the relatively small mass of stru-vite produced and the amount of handling and process lossesover the three months of this study are taken into account, therecovery was higher than expected. In a full-scale crystalliza-tion installation, it is expected that the losses would represent asmaller fraction of the produced struvite because of the largerscale involved.

Supersaturation ratioFigures 7 and 8 show the feed, in-reactor, and effluent SSRs

for reactors A and B, respectively. The most notable difference

between the SSRs in Figs. 7 and 8 are in the influent SSRs.This is because of the lower targeted recovery in reactor B.Because reactor B was operated at a lower pH, while the sameMg:PO4-P molar ratio as reactor A was maintained, the influentSSR was lower in reactor B. This allowed more supernatant tobe pumped through the reactor at a lower recycle ratio, whilethe same struvite crystal growth rate or struvite loading rate wasmaintained. Overall, by maintaining relatively constant struviteloading rates in both reactors, the in-reactor SSRs and the ef-fluent SSRs remained similar in both reactors.

Struvite product characteristicsThree major characteristics of the harvested struvite crystals

were investigated: size, apparent density, and chemical compo-sition. The crystals were also examined microscopically, bothby optical microscope and by scanning electron microscope(SEM).

Struvite crystal sizeIn general, the size of the harvested crystals continuously

increased during this study, even after several complete reactorvolumes had been harvested. Figure 9 shows the mean crystaldiameter of each harvest from both reactors. With time, thecrystals grew stronger and larger. One of the most importantfactors in determining the final diameter of the crystals was theirstructural strength. The early crystals tended to break easilyduring the harvesting, drying, and sieving operations, whereasthe crystals harvested late in this study did not tend to breakmuch during the subsequent handling processes.

These results imply that several complete CRTs must elapsebefore a steady-state crystal size will be reached, as suggestedby Takiyama et al. (1997). In fact, the mean crystal diametersshown in Fig. 9 do not appear to be approaching their steady-state values, even after 2 months of operation. Similar resultswere reported byAbe (1995). Further, longer term studies wouldbe necessary to determine the final steady-state size the struvitecrystals can be expected to reach. From the data collected inthis study, the mean crystal diameter grew by an average of0.016 mm per day.

Struvite crystal bulk densityAlthough the density of the crystals was not measured di-

rectly in this study, an interesting trend was observed in themass of crystals removed from the harvest section of the reac-tor with time. For each harvest, the flow through the reactor wasstopped and the fluidized crystals were allowed to settle beforethe harvest zone was isolated using ball valves and the crystalswere withdrawn from the reactor. In this manner, a constantvolume of approximately 1.1 L of crystals was harvested eachtime. It was noticed throughout the study that the crystal massharvested from this volume consistently increased with time.Figure 10 shows this trend.

The first harvests in the study had a total mass of approxi-mately 200 g, whereas the final harvests had a mass of approx-imately 600 g. This trend did not appear to reach a maximumby the end of this study. In general, the crystal mass harvested

© 2005 NRC Canada

Britton et al. 273

Fig. 7. Reactor A supersaturation ratios.

0

5

10

15

20

25

Date

SS

ratio

Inlet SS ratio In reactor SS ratio Effluent SS ratio

Fig. 8. Reactor B supersaturation ratios.

0

5

10

15

20

25

Date

SS

ratio

Inlet SS ratio In reactor SS ratio Effluent SS ratio

Fig. 9. Mean crystal diameter of struvite crystals harvested.

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

2

Date

Me

an

ha

rve

ste

dcry

sta

lsiz

e(m

m)

Reactor A Reactor B

© 2005 NRC Canada

274 J. Environ. Eng. Sci. Vol. 4, 2005

Fig. 10. Harvested crystal mass from the 1.1-L harvest zone.

0

100

200

300

400

500

600

700

Date

Ma

ss

ha

rve

ste

dfr

om

1.1

L(g

)

Reactor A Reactor B

Table 2. Average results of crystal composition analysis.

% Composition bymass

Mean Standarddeviation

Theoretical valuefor pure struvite

Mg 9.9 0.4 9.9N 5.6 0.3 5.7P 12.8 0.6 12.6Struvite (estimated) 99.8 4.0 100.0

increased by approximately 7 g per day of reactor operation.This increase in mass harvested coincided with the increase indiameter of the harvested crystals. Visual analysis of the crys-tals indicates that this increased bulk density of the crystalswas probably due to the change in shape of the crystals over thecourse of the study, from friable plate-like aggregates to harder,rounded crystal roses.

Struvite crystal compositionTo verify the composition of the crystals grown in this study,

29 samples were analyzed. Each size fraction (<0.5, 0.5–1,1–2, and >2 mm) of harvests, from three dates from each re-actor, were analyzed. Table 2 shows the average crystal com-positions compared with the expected theoretical compositionof pure struvite and the estimated struvite content of the crys-tals. The estimated struvite content was calculated by averag-ing the ratios of measured to theoretical composition of Mg,NH4-N, and PO4-P for each sample. The crystal samples werealso analyzed for content of Fe, Al, K, and Ca. Based on somepreliminary work, these were thought to be the most likely met-als that would be found in the struvite crystals as impurities.Table 3 shows the average results. The main impurity found inthe crystals was calcium, and it was present only at an averageof 0.5% by weight. Based on this analysis, the produced crystalswere classified as essentially pure struvite.

Table 3. Struvite crystal impurity content.

% Content by weight Mean Standard deviation

Ca 0.49 0.32K 0.04 0.01Fe 0.03 0.02Al * *

*Most Al analyses were below the detection limit of themethod used.

Microscope and scanning electron microscope crystalexamination

To better understand reasons for the observed changes in har-vested crystal sizes and densities, samples of the crystals werestudied, first under an optical microscope at 40× magnificationand then by SEM. A preliminary optical microscope observa-tion was used to screen samples and find general trends to selectrepresentative samples for further analysis by SEM.

In general, the appearance of the crystals changed signifi-cantly with time. The early crystals appeared to be a loose ag-gregation of plate-like crystals, which explains why the crystalstended to break apart during the drying and screening processes.Over the course of the study, the crystals grew more roundedand more solidly aggregated. The crystals from the final har-vests were very rounded and appeared to be monolithic undermicroscope examination at 40×.

Figure 11 shows SEM images of crystals that were retainedon a 1-mm sieve, but passed a 2-mm sieve. Images (a), (c), and(e) are of crystals harvested from reactor A, while images (b),(d), and (f) are from reactor B. The bar in the bottom right ofeach image represents 1 mm. Images (a) and (b) are of crys-tals harvested in October, images (c) and (d) are of crystalsharvested in November, and images (e) and (f) are of crystalsharvested in December.

© 2005 NRC Canada

Britton et al. 275

Fig. 11. Scanning electron microscope (SEM) images of crystals retained on a 1-mm sieve, but passing a 2-mm sieve at 50×magnification: (a) harvested 28 October 2001, from reactor A, (b) harvested 17 October from reactor B, (c) harvested 18 November fromreactor A, (d) harvested 18 November from reactor B (45×), (e) harvested 11 December from reactor A, and (f) harvested 12 Decemberfrom reactor B.

The progression in the crystal morphology is quite striking.The crystals appear to progress from loosely aggregated in Oc-tober to tightly packed and more solidly bound pellets in De-cember, especially in reactor B. The crystal from December inreactor A appears to be similar to the crystal from Novemberin reactor B. The outside faces of these crystals (November inreactor B and December in reactor A) appear to be similar tothe crystal from December in reactor B; however, their coresappear to be weaker. This may have caused the clumps of hardsurface crystals to break away, either because of turbulence andimpacts in the reactor or abrasion during drying and sievingoperations. These images can be used to explain the differencein the bulk densities of the harvested crystals, as was discussedpreviously. The later crystals appear to be much rounder and

more filled in, thus allowing the crystals to pack in more tightlyto the harvest zone, when flow through the reactor was stopped.

In Fig. 12 crystals from the four size fractions collected on11 December from reactor B are shown. Close inspection ofFig. 12c reveals that the individual surfaces on the inside of thecrystal show an orthorhombic shape and are growing from thecentre of the crystal outwards. This could lead to the conclusionthat the stripes observed on the surface of the whole crystalsmay in fact be the tips of these orthorhombic crystals, whichare rounded off by the abrasion in the fluidized bed and in thesieving process.

The stripes described above can be more clearly seen un-der higher magnification, as shown in Fig. 13. Under 300×magnification, the surface no longer appears as smooth as un-

© 2005 NRC Canada

276 J. Environ. Eng. Sci. Vol. 4, 2005

Fig. 12. Scanning electron microscope (SEM) images of crystals harvested from reactor B on 11 December 2001: (a) crystal retained ona 2-mm sieve (45×), (b) crystal retained on a 1-mm sieve (50×), (c) crystals retained on a 0.5-mm sieve (45×), and (d) crystals passinga 0.5-mm sieve (45×).

Fig. 13. Scanning electron microscope (SEM) images of a crystal retained on a 1-mm sieve, harvested from reactor B on 11 December2001: (a) 50×, (b) 300×, and (c) 4000×.

© 2005 NRC Canada

Britton et al. 277

der 50× magnification, and it appears that the crystal is madeup of tightly agglomerated, smaller, brick-like crystals. Under4000× magnification, the surface of each individual crystal tipbecomes more resolved, and it becomes apparent that thesecrystal surfaces are covered with cracks and fissures (Fig. 13).

The crystallography of the recovered crystals was not ana-lyzed in depth but simply observed visually. Because the crys-tals were grown under continually varying conditions, it is im-possible to determine the exact cause of the changes in the ap-pearance of the crystals. It would also be impossible to predictthe appearance of crystals from other recovery systems basedon this study, especially considering that the crystals grownin this study were very different from those grown in otherstudies using synthetic supernatant in similar reactors (Adnan2002; Adnan et al. 2003b). Further studies, under tightly con-trolled conditions, would probably be necessary to determinethe cause-and-effect relationships leading to the final shape andsize of the crystals produced.

Conclusions

Based on the results obtained from this pilot-scale study ofstruvite recovery from a full-scale anaerobic digester, the fol-lowing conclusions can be drawn:

• The pilot-scale struvite recovery reactor developed at UBCwas effective in removing orthophosphate from anaerobicdigester supernatant stream under controlled conditions andproduced a product consisting of nearly pure struvite.

• By controlling the operating pH of the reactors or the inletSSR, the percentage removal of orthophosphate in the reactorvaried between 42% and 91%. This study’s target orthophos-phate reduction of at least 70% was easily and consistentlyachieved.

• During the course of this study, 90% to 91% of the removedphosphorus was recovered after subsequent harvesting, dry-ing, and screening operations.

• The size and hardness of the struvite crystals appear to havebeen affected by several factors, including CRT, SSR, andelapsed time from reactor start-up. It was not possible to de-termine the exact effect of these parameters separately fromthis study, because they were not varied independently.

Acknowledgements

This research project was supported financially by BritishColumbia Hydro and Power Authority; Stantec ConsultantsLtd.; the City of Penticton, British Columbia; and the NaturalSciences and Engineering Research Council of Canada. The au-thors would like to especially thank the operators and technicalstaff at the City of Penticton AWWTP, as well as the laboratorytechnicians and staff at the UBC Environmental Laboratory fortheir support and assistance throughout this study.

References

Abe, S. 1995. Phosphate removal from dewatering filtrate by MAPprocess at Seibu treatment plant in Fukuoka City. Sewage WorksJpn. 43: 59–64.

Adnan,A. 2002. Pilot-scale study of phosphorus recovery through stru-vite crystallization. M.Sc. thesis, Department of Civil Engineering,The University of British Columbia, Vancouver, B.C.

Adnan, A., Mavinic, D.S., and Koch, F.A. 2003a. Pilot-scale study ofphosphorus recovery through struvite crystallization — examiningthe process feasibility. J. Environ. Eng. Sci. 2: 315–324.

Adnan, A., Koch, F.A., and Mavinic, D.S. 2003b. Pilot-scale study ofphosphorus recovery through struvite crystallization — II:Applyingin-reactor supersaturation ratio as a process control parameter. J.Environ. Eng. Sci. 5: 473–483.

APHA/AWWA/WEF. 1995. Standard met hods for the examination ofwater and wastewater. 19th ed. American Public Health Associa-tion, American Water Works Association, and Water EnvironmentFederation. American Public Health Association, Washington, D.C.

Britton, A.T. 2002. Pilot-scale struvite recovery trials from a full-scaledigester supernatant at the City of Penticton advanced wastewatertreatment plant. M.Sc. thesis, Department of Civil Engineering, TheUniversity of British Columbia, Vancouver, B.C.

Dastur, M.B. 2001. Investigation into the factors affecting controlledstruvite crystallization at the bench-scale. M.Sc. thesis, The Uni-versity of British Columbia, Vancouver, B.C.

Doyle, J.D., Phillip, R., Churchley, J., and Parsons, S.A. 2000.Analysisof struvite precipitation in real and synthetic liquors. Trans. Inst.Chem. Eng., Part B, 78: 480–488.

Driver, J., Lijmbach, D., and Steen, I. 1999. Why recover phosphorusfor recycling and how? Environ. Technol. 20: 651–662.

Jardin, N., and Popel, H.J. 1996. Behaviour of waste activated sludgefrom enhanced biological phosphorus removal during sludge treat-ment. Water Environ. Res. 68: 965–973.

Jardin, N., and Popel, H.J. 2001. Refixation of phosphates releasedduring bio-P sludge handling as struvite or aluminum phosphate.Environ. Technol. 22: 1253–1262.

Mavinic, D.S., Koch, F.A., Hall, E.R., Abraham, K., and Niedbala, D.1998. Anaerobic co-digestion of combined sludges from a biologi-cal phosphorus removal process. Environ. Technol. 19: 35–44.

Munch, E.V., and Barr, K. 2001. Controlled struvite crystallization forremoving phosphorus from anaerobic digester sidestreams. WaterRes. 35: 151–159.

Niedbala, D. 1995. Pilot-scale studies of the anaerobic digestion ofcombined wastewater sludges and mitigation of phosphorus release.M.Sc. thesis, The University of British Columbia, Vancouver, B.C.

Ohlinger, K.N. 1999. Struvite controls in anaerobic digestion and post-digestion wastewater treatment processes. Ph.D. thesis, Universityof California, Davis, Calif.

Snoeyink, V., and Jenkins, D. 1980. Water chemistry. John Wiley &Sons, New York.

Takiyama, H., Yamauchy, H., Matsuoka, M. 1997. Effects of seedingon start-up operation of a continuous crystallizer. In Separation andpurification by crystallization. ACS Symp. Ser. 667: 172–186.

Ueno, Y., and Fujii, M. 2001. Three years experience of operating andselling recovered struvite from full-scale plant. Environ. Technol.22: 1373–1381.

Woods, N.C., Sock, S.M., and Daigger, G.T. 1999. Phosphorus recov-ery technology modeling and feasibility evaluation for municipalwastewater treatment plants. Environ. Technol. 20: 663–680.

© 2005 NRC Canada